_gpu_dxg_r4_recon.nx source

1// _gpu_dxg_r4_recon.nx -- SOVEREIGN-GPU ladder R4 RECON (create a GPU CONTEXT on the WDDM device). 2// 3// R3 made a real WDDM device. R4 (command-submission sub-arc) starts with a CONTEXT -- the queue GPU 4// work is submitted on. WSL2 typically routes this through LX_DXCREATECONTEXTVIRTUAL (paravirt path). 5// No uapi header on this kernel, so discover the ABI live: sweep BOTH nr=0x03 (CREATECONTEXT) and 6// nr=0x04 (CREATECONTEXTVIRTUAL) across candidate sizes with device@0 + node_ordinal/engine/hint = 0 and 7// a NULL priv_drv_data(size 0). ret==0 marks the real size; scan for a 0x4000xxxx OUT context handle. 8// A non-ENOTTY error (e.g. -EINVAL/-22) at a given size means that nr+size IS the recognized ABI but the 9// INPUT fields need more than zeros (then R4a needs a focused session). raw syscalls only. No log writes. 10// license_tier: ORIGINAL 11import "nx_syscalls.nx" 12 13const ENUM2_CODE: i64 = 0xC0104714 14const QAI_CODE: i64 = 0xC0184709 15const OAFL_CODE: i64 = 0xC00C4701 16const CDEV_CODE: i64 = 0xC0404702 // LX_DXCREATEDEVICE (R3) 17 18func p(s: *u8) -> i64 { var nn: i64=0; while s[nn]!=(0 as u8){nn=nn+1} sys_write(1,s,nn); return 0 } 19func n(v: i64) -> i64 { let bb: *u8=sys_mmap(28); var m: i64=v; if m<0{m=0-m;sys_write(1,"-" as *u8,1)}; let t: *u8=sys_mmap(28); var k: i64=0; if m==0{t[0]=48;k=1}; while m>0{t[k]=(48+(m%10)) as u8;m=m/10;k=k+1}; var i: i64=0; while i<k{bb[i]=t[k-1-i];i=i+1}; sys_write(1,bb,k); return 0 } 20func x(v: i64) -> i64 { p("0x" as *u8); let bb:*u8=sys_mmap(20); var k:i64=0; var m:i64=v; if m==0{bb[0]=48;k=1}; while m>0{ let d:i64=m&15; if d<10{bb[k]=(48+d) as u8}else{bb[k]=(87+d) as u8}; m=(m>>4); k=k+1 } var i:i64=0; let o:*u8=sys_mmap(20); while i<k{o[i]=bb[k-1-i];i=i+1} sys_write(1,o,k); return 0 } 21func iowr(nr: i64, size: i64) -> i64 { return (3 << 30) | ((size & 0x3fff) << 16) | (0x47 << 8) | (nr & 0xff) } 22func rd32(buf: *u8, off: i64) -> i64 { return (buf[off] as i64)|((buf[off+1] as i64)<<8)|((buf[off+2] as i64)<<16)|((buf[off+3] as i64)<<24) } 23 24func query_type(fd: i64, handle: i64, qtype: i64, psize: i64, outv: *i64) -> i64 { 25 let priv: *u8 = sys_mmap(64); var zz: i64 = 0; while zz < 64 { priv[zz] = 0 as u8; zz = zz + 1 } 26 let req: *u8 = sys_mmap(64); var y: i64 = 0; while y < 64 { req[y] = 0 as u8; y = y + 1 } 27 let h: *i64 = (req as i64 + 0) as *i64; h[0] = (handle & 0xffffffff) | ((qtype & 0xffffffff) << 32) 28 let pd: *i64 = (req as i64 + 8) as *i64; pd[0] = priv as i64 29 let pds: *i64 = (req as i64 + 16) as *i64; pds[0] = psize & 0xffffffff 30 let ret: i64 = sys_ioctl(fd, QAI_CODE, req as i64); outv[0] = rd32(priv, 0); return ret 31} 32func open_from_luid(fd: i64, luid_lo: i64, luid_hi: i64, outh: *i64) -> i64 { 33 let req: *u8 = sys_mmap(64); var y: i64 = 0; while y < 64 { req[y] = 0 as u8; y = y + 1 } 34 let lo: *i64 = (req as i64 + 0) as *i64; lo[0] = (luid_lo & 0xffffffff) | ((luid_hi & 0xffffffff) << 32) 35 let ret: i64 = sys_ioctl(fd, OAFL_CODE, req as i64); outh[0] = rd32(req, 8); return ret 36} 37func create_device(fd: i64, adapter: i64, outd: *i64) -> i64 { 38 let buf: *u8 = sys_mmap(512); var bz: i64 = 0; while bz < 512 { buf[bz] = 0 as u8; bz = bz + 1 } 39 let a0: *i64 = (buf as i64) as *i64; a0[0] = adapter & 0xffffffff 40 let ret: i64 = sys_ioctl(fd, CDEV_CODE, buf as i64); outd[0] = rd32(buf, 12); return ret 41} 42 43// sweep candidate sizes for a given nr, with device@0 and the rest zero (NULL priv data). 44func sweep(fd: i64, nr: i64, device: i64) -> i64 { 45 let sizes: *i64 = sys_mmap(256) 46 sizes[0]=24; sizes[1]=32; sizes[2]=40; sizes[3]=48; sizes[4]=56; sizes[5]=64; sizes[6]=72; sizes[7]=80; sizes[8]=96; sizes[9]=128; sizes[10]=0 47 var si: i64 = 0 48 while sizes[si] != 0 { 49 let sz: i64 = sizes[si] 50 let buf: *u8 = sys_mmap(512); var bz: i64 = 0; while bz < 512 { buf[bz] = 0 as u8; bz = bz + 1 } 51 let d0: *i64 = (buf as i64) as *i64; d0[0] = device & 0xffffffff // device @0 (in) 52 let code: i64 = iowr(nr, sz) 53 let ret: i64 = sys_ioctl(fd, code, buf as i64) 54 p(" nr=" as *u8); x(nr); p(" size=" as *u8); n(sz); p(" code=" as *u8); x(code); p(" ret=" as *u8); n(ret) 55 if ret != 0 - 25 { // anything other than -ENOTTY means this nr+size is a RECOGNIZED ioctl 56 var off: i64 = 4; var fo: i64 = 0 - 1; var fv: i64 = 0 57 while off + 4 <= sz { let w: i64 = rd32(buf, off); if (w & 0xffff0000) == 0x40000000 { if fo < 0 { fo = off; fv = w } } off = off + 4 } 58 p(" <- RECOGNIZED (non-ENOTTY). 0x4000-handle scan: off=" as *u8); n(fo); p(" val=" as *u8); x(fv) 59 } 60 p("\n" as *u8) 61 si = si + 1 62 } 63 return 0 64} 65 66func main() -> i64 { 67 p("=== R4 RECON: discover CREATECONTEXT/CREATECONTEXTVIRTUAL ABI on the 5080 WDDM device ===\n" as *u8) 68 let fd: i64 = sys_openat_rd("/dev/dxg" as *u8) 69 if fd < 0 { p("open(/dev/dxg) failed\n" as *u8); sys_exit(1); return 1 } 70 let ainfo: *u8 = sys_mmap(4096); var z: i64 = 0; while z < 4096 { ainfo[z] = 0 as u8; z = z + 1 } 71 let ereq: *u8 = sys_mmap(64); ereq[0] = 8 as u8; let r8: *i64 = (ereq as i64 + 8) as *i64; r8[0] = ainfo as i64 72 sys_ioctl(fd, ENUM2_CODE, ereq as i64) 73 let nc: i64 = rd32(ereq, 0) 74 let v: *i64 = sys_mmap(16); var disc_lo: i64 = 0; var disc_hi: i64 = 0 75 var ai: i64 = 0 76 while ai < nc { 77 let base: i64 = ai * 20; let eh: i64 = rd32(ainfo, base) 78 let qret: i64 = query_type(fd, eh, 15, 4, v); let t: i64 = v[0] 79 if qret == 0 { if ((t>>4)&1) == 1 { if ((t>>2)&1) == 0 { disc_lo = rd32(ainfo, base+4); disc_hi = rd32(ainfo, base+8) } } } 80 ai = ai + 1 81 } 82 let ah: *i64 = sys_mmap(16); open_from_luid(fd, disc_lo, disc_hi, ah) 83 let dh: *i64 = sys_mmap(16); let dr: i64 = create_device(fd, ah[0], dh) 84 p(" adapter=" as *u8); x(ah[0]); p(" create_device ret=" as *u8); n(dr); p(" device=" as *u8); x(dh[0]); p("\n" as *u8) 85 86 p("\n-- sweep nr=0x03 (LX_DXCREATECONTEXT) --\n" as *u8); sweep(fd, 3, dh[0]) 87 p("\n-- sweep nr=0x04 (LX_DXCREATECONTEXTVIRTUAL) --\n" as *u8); sweep(fd, 4, dh[0]) 88 89 sys_close(fd) 90 p("=== R4 RECON DONE ===\n" as *u8) 91 return 0 92}